[Pool] question on behavior read from chart

[Ask Bjørn Hansen]

On Sep 1, 2010, at 7:38, Chuck wrote:

> is this something i can/should correct in the ntp time1 "fudge" or is it best left alone?
> at a consistant -2ms i wonder if it is even 'critical'.

You are overestimating the accuracy of the monitoring server.

If someone analyzed all the scores across all the servers and decided that the monitoring server consistently is a few ms off, I wouldn't be too alarmed.

Again, the goal of the monitoring server is to detect servers that are down or serving "bad time" where "bad time" is defined as "something that'd be bad for an sntp client".  An SNTP client should by definition (in my opinion) be more than happy with something that's within a few hundred ms.


   - ask